RapidMiner

Error SparkRM

Regular Contributor

Error SparkRM

Hello Smiley Happy

I am using sparkRM to create a model for big data, but this error always appere Smiley Sad

Can you help me please

 

Capture d’écran 2017-06-28 à 12.32.17.png

 

Attachments

9 REPLIES
Contributor

Re: Error SparkRM

Hello,

 

I assume that you are using a virtual machine? In this case, try increasing the total amount of memory for the VM. Then you can also edit the parameter yarn.scheduler.maximum-allocation-mb in the "Advanced Connection Properties" of the set up/configured Radoop connection.

Saw a similar issue a while back and by playing around with memory/ressource parameters I had it working. Can't remember exactly now...

 

Regards

 

RMStaff

Re: Error SparkRM

Hi,

 

I also assume it is a VM with small memory resources (quickstart VM?).

The simplest quickfix is probably to change the Radoop connection setting "Spark Resource Allocation Policy" from "Static, Heuristic Configuration" to "Static, Default Configuration".

 
Screen Shot 2017-06-28 at 22.17.14.png

 

This results in starting small containers (it is only suitable for single node test clusters).

 

Best,

Peter

  

 

 

Regular Contributor

Re: Error SparkRM

Thank you so much Smiley Happy your answer was really helpful

but an other problem appear Smiley Sad after resolving the first one Smiley Sad

 

can you help me please Smiley SadCapture d’écran 2017-07-05 à 08.39.43.png

RMStaff

Re: Error SparkRM

Do you have access to the Resource Manager interface? Typically, you can access it via a web browser at <hostname>:8088. You should see a job with the Name "Single Process Pushdown" there with the corresponding StartTime. If you click on its ID, it brings up a page, where there should be more details next to Diagnostics (somewhere in the middle).

diagnostics.png

 

 Can you share it please?

 

Peter

 

 

Regular Contributor

Re: Error SparkRM

thank you for replying Smiley Happy

 

 

 

Capture d’écran 2017-07-05 à 13.23.03.png

RMStaff

Re: Error SparkRM

Oh, almost...

Looks like you will need to go further with clicking on one of the logs link. The cause may be there, or following the here links may reveal more.

Regular Contributor

Re: Error SparkRM

[ Edited ]

Thank you Smiley Happy

but I can't open the link you put it in the message (here links..)

Regular Contributor

Re: Error SparkRM

hello Smiley Happy

I saw the log file in cloudera manager and I found this error:

[HiveServer2-Background-Pool: Thread-504]: Table radoop__tmp_lola_1499279743370_ruyrrcu not found: default.radoop__tmp_lola_1499279743370_ruyrrcu table not found

So the cloudera comunity suggest that "the users using RM can't create tables under default. Or is creating them but under a different DB but RM Radoop is looking for them under default".

 

So can you tell me please how to solve this in Rapidminer?

RMStaff

Re: Error SparkRM

Hi,

 

these "table not found" error messages in the log do not mean anything, they are normal operation. If you run "drop table if exists tmp_xyz" in Hive, and the table does not exist, you get this log entry despite the "if exists" part, but no error appears on the client side.

 

With the logs link I meant this link below the Diagnostics:

logs.png

 Peter